Sam Altman: Intelligence Got 100x…

1. Key Themes

Theme 1: The 10x/Year Intelligence Deflation Curve Is the Foundational Investment Thesis

AI inference costs are falling at a rate that dwarfs Moore's Law, making today's unit economics obsolete almost immediately — and rewarding those who plan against the future price, not the current one.

"Our observation in our field is that the price for a given level of intelligence, once it's achieved, falls by about 10x per year. So a hundred — a hundred x in the last two years."

Theme 2: AI Is Already Displacing Knowledge Work at Scale — One Feature at a Time

Altman contends that a single OpenAI product, Deep Research, already handles a material slice of the economy's cognitive output. Knowledge workers — analysts, researchers, consultants — are the first wave of disruption, well ahead of physical labor.

"My vibes-based estimate is that [Deep Research] does about 5% of all tasks in the economy today, one feature, 5% of all the tasks in the economy today."

The specific task list he names — summarizing entire research fields, running full financial analyses, producing consultant-quality reports — maps directly onto junior knowledge worker roles.

Theme 3: Sovereign AI Infrastructure Is Becoming a National Budget Priority

Governments are no longer treating AI compute as a research grant — they are negotiating to own infrastructure at scale. This represents a new, durable category of public-sector demand that Altman is actively courting.

"I was pleasantly surprised on this trip... There's some governments that are ready to like buy big pieces of AI infrastructure."

For context, the UK's current national supercomputer commitment is ~£800M — a fraction of what Altman says governments are now requesting directly from OpenAI.

Theme 4: Physical Infrastructure (Power + Permits) Is Now a Strategic Moat

The ability to train and deploy frontier AI is gated not just by capital or talent, but by permitting speed and energy access. Altman frames regulatory environment as a direct variable in competitive positioning.

"President Trump has such a different opinion on building things, and permits, and power, and manufacturing in the US."

He explicitly called the Biden administration "not easy on the infrastructure front," naming permitting as the specific bottleneck — making jurisdiction selection a first-order strategic decision for any infrastructure-dependent company.

Theme 5: The $500B Stargate Is a Stepping Stone to a $5 Trillion Build

Altman's capital ambitions are not incremental — he is signaling that the Stargate project is a proof-of-concept for an order-of-magnitude larger infrastructure program, including a Stargate Europe already in early conversations.

"Stargate is a $500 billion project to build a very large training inference system. It sounds crazy big now. I bet it won't sound that big in a few years. If we get to do this again, you'll be raising $5 trillion for a cluster."

He also signals opportunistic M&A logic in the event of an infrastructure bubble correction: he would "happily buy someone else's overbuilt infrastructure at 10 cents on the dollar."

2. Contrarian Perspectives

Perspective 1: DeepSeek Was Not a Research Threat — Markets Overreacted

While DeepSeek's cheap, capable model triggered ~$1 trillion in market value erasure across AI-adjacent equities, Altman says it was a non-event internally at OpenAI's research level. The market's fear response was disconnected from how frontier labs actually assessed the competitive signal.

"[DeepSeek was] not a big research update internally."

He credited DeepSeek with two smart product choices — showing the model's chain of thought and offering a generous free tier — but pointedly did not elevate it to a research breakthrough. This implies the market panic was a sentiment event, not a fundamental one, and potentially a buying opportunity in AI infrastructure that was treated as a risk-off signal.

Perspective 2: Tech's Political Alignment with Trump Is Transactional, Not Ideological — and That's Deliberate

The conventional read of Silicon Valley's pivot toward Trump is ideological or social. Altman's framing is colder: founders are buying permitting speed and infrastructure access, not endorsing a political platform. Donations to inauguration funds and presidential libraries are, in his telling, inputs to a construction strategy.

"The word that came to mind for the last administration was hostile. I think that's a little bit too strong, but they were not friendly to tech or business. It's a very welcome breath of fresh air."

He adds that the upside is concrete — "rebuilding semiconductor fabrication, robotic factories for data centers, and new energy generation inside the US" — categories the US had largely ceded. The implication for investors: political risk in AI infrastructure is now a function of permitting velocity, not regulatory ideology.

Perspective 3: The Bottleneck Will Shift From Model Capability to Human Imagination

Altman's projection is that the constraint on AI value creation will stop being what the model can do, and become what humans can think to ask for. This inverts the conventional framing that capability is the limiting factor.

"People like a lot of people who were maybe even recent AI skeptics were saying things like, 'I can now do things that would have taken me many days or even weeks of work. AI can do it in 20 minutes and it can do a bunch of them in parallel.'"

Run the 10x/year curve forward two years and the ceiling is no longer model performance — it is prompt quality, workflow design, and the human ability to decompose complex problems into AI-executable tasks. This argues for investing in the interface and orchestration layer, not just the model layer.

3. Companies Identified

OpenAI

Description: Leading AI lab, creator of ChatGPT and Deep Research
Why mentioned: Central subject; Altman discusses pricing trajectory, Deep Research's market impact, Stargate infrastructure ambitions, and the for-profit structural shift
Quote: "Stargate is a $500 billion project to build a very large training inference system."

DeepSeek

Description: Chinese AI lab that released a high-capability, low-cost model
Why mentioned: Used as a case study for how markets overreacted to a competitor that Altman says did not represent a genuine research update
Quote: "[DeepSeek was] not a big research update internally" — though Altman credited "a couple of smart product choices: showing the model's chain of thought, and a generous free tier."

SoftBank

Description: Japanese conglomerate and investment firm
Why mentioned: Named as lead investor in the $40B OpenAI raise that is a component of the larger Stargate build
Quote: "OpenAI is currently raising $40 billion specifically for Stargate, led by SoftBank."

Anthropic

Description: AI safety-focused lab and OpenAI competitor
Why mentioned: Referenced as independent evidence of the 10x/year cost deflation curve, having cut the price of frontier intelligence in half
Quote: Referenced in the article as "Anthropic cutting the price of frontier intelligence in half" — corroborating Altman's cost curve claim.

4. People Identified

Sam Altman

Description: CEO of OpenAI, co-founder
Why mentioned: Primary interview subject; source of all 10 core takeaways on AI pricing, infrastructure investment, geopolitics, and competitive dynamics
Quote: "The price for a given level of intelligence, once it's achieved, falls by about 10x per year."

Elon Musk

Description: CEO of Tesla and SpaceX, co-founder of OpenAI, founder of xAI
Why mentioned: Former friend and co-founder of OpenAI, now an active legal adversary; used to illustrate OpenAI's for-profit structural evolution and the personal cost of the split
Quote: "Do you miss him as a friend? No, I don't. I think he's really changed."

Paul Graham

Description: Co-founder of Y Combinator, essayist and startup philosopher
Why mentioned: His decade-old characterization of Altman — "extremely good at becoming powerful" — closes the interview and goes unanswered, raising questions about intention vs. outcome in Altman's rise
Quote: "I can't argue that I ended up in a fairly influential position. I don't know how to square those." (Altman's response to the Graham quote)

Katie Prescott & Danny Fortson

Description: Journalists at The Times (UK)
Why mentioned: Conducted the Times Tech Podcast interview in London that is the source material for this entire article
Quote: Referenced as the interviewers who "put the comparison directly" to Altman on Biden vs. Trump

5. Operating Insights

Insight 1: Re-Audit Your Team's Task List Against the 5% Benchmark — Now

Altman's claim that one feature already handles ~5% of all economic tasks is a direct prompt for operators to audit which of their team's workflows already fall inside that coverage — and treat it as a floor, not a ceiling.

"My vibes-based estimate is that [Deep Research] does about 5% of all tasks in the economy today, one feature, 5% of all the tasks in the economy today."

Tactical application: Map your knowledge workers' weekly tasks against Deep Research's named capabilities (research summarization, financial analysis, consultant-quality drafting, comparative product analysis). Any task that overlaps is an immediate candidate for AI-first workflow redesign.

Insight 2: Price Your Product Roadmap Against Tomorrow's Inference Cost, Not Today's

The 10x/year cost curve means that products which are currently too expensive to build profitably may become viable within 12–24 months. Roadmapping against today's cost locks you into underambitious product decisions.

"The price for a given level of intelligence, once it's achieved, falls by about 10x per year. So a hundred x in the last two years."

Tactical application: Build financial models with a "cost curve scenario" column that assumes 10x annual cost reduction in AI inference. Products that break even at 10x lower cost today should be on the active roadmap, not the wish list.

Insight 3: Institutionalize Competitive Paranoia as a Cultural Practice

Altman's answer to the "Napster risk" question is not a strategic framework — it is a daily emotional discipline. For operators, this translates to building competitive review cadences into routine operations, not just quarterly strategy sessions.

"Of course I'm worried about it. You got to wake up every day — the only way to not have that happen is to wake up every day worried about it."

6. Overlooked Insights

Insight 1: Altman Would Deliberately Buy Distressed AI Infrastructure at a Discount

Buried in the Stargate discussion is a strategic signal that often gets skipped: Altman openly anticipates a boom-bust cycle in AI infrastructure buildout and has a pre-formed plan to acquire overbuilt capacity at steep discounts when corrections come. This is not a defensive hedge — it is an offensive M&A posture.

He "says he'd happily buy someone else's overbuilt infrastructure at 10 cents on the dollar when that correction comes."

For investors watching data center REITs, GPU leasing companies, and hyperscaler capex, this is a signal that the smart money views a correction as a buying event, not a structural bear case.

Insight 2: Sovereign AI Access May Default to Cloud Services — Not Ownership

Altman's fallback offer to governments that cannot afford to own large-scale AI infrastructure is to purchase AI capability as a cloud service. This quietly establishes a new government-facing revenue tier — and a geopolitical dependency model — that has significant implications for non-US sovereigns who want AI capability without the capital or political will to own it.

"His fallback for governments that won't write a check that size is direct: buy AI as a cloud service instead of owning the infrastructure, on the bet that the world stays open enough to sell it to them."

The qualifier — "on the bet that the world stays open enough" — is doing significant work here. It implies Altman himself views deglobalization as a non-trivial risk to this model, which is an underreported geopolitical bet embedded in OpenAI's go-to-market strategy.